biubug6 / Pytorch_Retinaface

Retinaface get 80.99% in widerface hard val using mobilenet0.25.
MIT License
2.63k stars 774 forks source link

SSH code is different from paper #110

Closed gaoruiyang closed 4 years ago

gaoruiyang commented 4 years ago
def forward(self, input):
    conv3X3 = self.conv3X3(input)

    conv5X5_1 = self.conv5X5_1(input)
    conv5X5 = self.conv5X5_2(conv5X5_1)

    conv7X7_2 = self.conv7X7_2(conv5X5_1)
    conv7X7 = self.conv7x7_3(conv7X7_2)

    out = torch.cat([conv3X3, conv5X5, conv7X7], dim=1)
    out = F.relu(out)
    return out

I think the code is different from paper. conv5X5_1 = self.conv5X5_1(input) should be changed to conv5X5_1 = self.conv5X5_1(conv3X3) Could you please tell me why?

gaoruiyang commented 4 years ago

Actually, I think the picture in paper at top of page 4 maybe wrong. SSH module is used just like the code in this git.